Using Reinforcement Learning to Optimize the Global and Local Crossing Number